Research Scientist: Real-time Multimodal Generative AI

Google DeepMind; Greater London; On-site; GBP 90,000 - 150,000; Full time

Job summary

Google DeepMind is seeking a Research Scientist to develop multimodal generative models and advance the frontier of world modeling research, focusing on diffusion models for real-time video generation. You will join our core modeling team to build capable, interactive models and collaborate with researchers and engineers to set up tight feedback loops across modeling and data efforts.

You will be involved in significant research, road-mapping, and engineering work to deliver world models, with

Qualifications

• PhD or equivalent in ML/CS/math or related field.
• 2 years publishing ML research or releasing software models.
• 2 years in diffusion/multimodal/generative video models.

Responsibilities

• Drive research and implementation of new architectures to build frontier generative models with real-time performance.
• Design, evaluate, and generate datasets to discover new model capabilities.
• Collaborate across capability workstreams to discover and scale optimal training recipes.
• Work across the training stack to land new generative and world model capabilities.
• Stay current with the latest research on diffusion models and generative video architectures.

Skills

ML research

Research publications

Diffusion models

Education

PhD or equivalent in ML/CS/math

Tools

JAX

CUDA

Job description

Google DeepMind is seeking a Research Scientist to develop multimodal generative models and advance the frontier of world modeling research, focusing on diffusion models for real-time video generation. You will join our core modeling team to build capable, interactive models and collaborate with researchers and engineers to set up tight feedback loops across modeling and data efforts.

You will be involved in significant research, road-mapping, and engineering work to deliver world models, with